Machine learning will never replace organic human creativity (unless a miracle of AI produces Ultron, Skynet, Chappie, or Johnny-5). In other words, creative marketers and charismatic sales people need not worry; machine learning algorithms aren’t going to put you out of a job any time soon.
Machine learning, as the common definition states, is the subfield of computer science that “gives computers the ability to learn without being explicitly programmed” (Arthur Samuel, 1959). The key word here, of course, being “explicitly.” In essence, machine learning makes it possible for an imperfect program to perfect itself. Imagine, for example, a very simple program that used a quiz to help you figure out what dog is best for you. The basic version would offer a finite number of results based on pre-established connections to the questions.
Machine learning capabilities, though, would take the questions, a database of dogs, and feedback from a final question such as, “Did you like this dog – yes or no?”, and gradually improve its responses. While the program might originally only be able to serve up dogs based on size (small, medium, or large), over time it could zero in on the right dog breed for each person.
Say Hello to Murphy
If you would like to see how machine learning works in practice, you should checkout a quirky Microsoft chat bot named Murphy.
Murphy’s learning parameters are based on “what if” questions. Its goal is to take the two aspects named in a “what if” question – “What if Abraham Lincoln had a longer beard?” – and put them together to produce a visual of the suggested scenario.
The results it produces are in no way well-crafted. Presently, the only formula that seems to work is, “What if [popular thing A] was [popular thing B]?” You can’t really ask it, “What if Vincent Van Gogh had painted the Mona Lisa?” (Editor: Though there are programs that could pull this off!]
For some queries, Murphy’s output can seem forced.
Some things, by chance or fortuitous positioning of existing images, mesh well.
And some things, well, some things just end up being delightfully weird.
Applications for Marketing and Sales
In the case of Murphy, machine learning has created an addictive piece of digital real-estate for Microsoft, as well as for the communications channels that support interactions with Murphy – channels that include Skype, Facebook Messenger, and Telegram (Kik and Slack channels coming soon!).
Speaking from personal experience, I find it extremely addictive to type in “what if” scenarios and get quirky results. The first reaction I have when something fun comes up is to share it with friends, further expanding the web of interactions around Murphy.
As shown in Aberdeen’s recent content marketing research, it is precisely this ability to effectively invite sharing, mentions, or external coverage that separates top-performing content marketers from their peers. Leading content marketers, as it turns out, are twice as effective at earning positive social media pickup or other media coverage than their underperforming peers (74% vs. 36%).
Social and earned media aside, machine learning could be used to create unique and targeted interactive content. While Murphy is certainly fun, imagine setting parameters in a widget enabled by machine learning that could generate ROI calculations for potential buyers.
Similarly, on-site search functions could use machine learning to recommend marketing content or sales collateral depending on what a visitor was looking for (or had been looking at). Finally, while marketers commonly create pre-defined lead nurturing paths or on-site, guided click streams, the refinement and optimization of these systems could be automated with machine learning.
In the world of sales, one need only consider sales performance management (SPM) tools to see the impact machine learning could have. SPM technology includes solutions that support the end-to-end data collection, analysis, workflow creation, and overall management of sales operations in order to systematically improve individual and team sales performance. As one example of the kinds of advantages SPM offers, Aberdeen research found that SPM users improve annual quota attainment at a 49% greater rate than non-users (8.67% vs. 5.83%).
What if you could use machine learning to improve the analysis of performance data and thus produce more efficient workflows? It would be like giving your SPM solution superpowers!
The possible applications of machine learning to every aspect of your business are literally unlimited. In the marketing and sales spaces alone, one can easily imagine how it could improve everything from usability and customer experience to overall marketing and sales performance.
In the end, the primary strength of machine learning isn’t that it can or should produce anything utterly new on its own. Instead, it’s strength lies in the ability to automate the improvement and refinement, over time, of processes and programs marketers and sales leaders create.
That being said, sometimes machine learning can produce things that need no improvement: